Posts

One day I may find time to finally make a blog, but for now I am using this space to collect and link to some of the (quite elaborate) threads I’ve made explaining some papers! (find me @AliciaCurth on Twitter or Bluesky!)

November 2024: Telescoping Approximation & Understanding Model Complexity in Deep Learning. From double descent to grokking, deep learning sometimes works in unpredictable ways… or does it? For NeurIPS24, we explored if & how statistics + smart linearisation can help us better understand & predict numerous odd deep learning phenomena — and learned a lot.. 🧵 Continued here. 📖 Corresponding paper.

October 2024: Benign Overfitting. When Double Descent & Benign Overfitting became a thing, I was a masters student in statistics — and so confused. I couldn’t reconcile what l had literally just learned about bias-variance&co with modern ML 😢 Here’s what I wish someone had told me then: 🧵 Continued here. 📖 Corresponding paper.

Feb 2024: Random Forests. Why do Random Forests perform so well off-the-shelf & appear essentially immune to overfitting?!? I’ve found the text-book answer “it’s just variance reduction 🤷🏼‍♀️” to be a bit too unspecific, so we investigate 🕵️‍♀️… 🧵 Continued here. 📖 Corresponding paper.

Nov 2023: Double Descent. Every StatML intro class covers complexity-error U-curves, so we asked ourselves whether the info from theses classes would be enough to explain double descent too? Our NeurIPS23 paper does a roundtrip of the Elements of statistical learning and answers “Yes”! 🧵 Continued here.📖 Corresponding paper.